Team members:
Mark Sucrese (VP of Marketing Sciences), with Kevin Ng, Ravi Nagabhyru, Ben Eubank, Felice Brezina, Raghavan Kirthivasan, Kevin Elwood, Ben McVay, and Wayne Townsend.
Country:
United States
Organization:
Epsilon DX Machine Learning
Description:
Epsilon DX is an organization that is focused on delivering value through; partnerships, engineering, creativity, strategy, software implementation, and best-in-class managed services. We work with our clients to help them take on the challenges of today, tomorrow, and beyond. We have deep expertise in partnerships like; Adobe, Salesforce, Dataiku, IBM, Microsoft, and Sitecore to create unique partnerships that bring value to our clients not seen by others.
Awards Categories:
Challenge:
We have been asked by a large US retailer to create a universal decision engine, that leverages modern machine learning technology, to create omnichannel personalized experiences that can scale the enterprise for digital and non-digital customer engagements.
Areas of focus for personalization are:
- Product and offer recommendations,
- Optimized content and messaging,
- Improved and targeted pricing,
- Ability to reduce fraud.
The brand wants to ensure that model development can be leveraged by both data scientists and business analysts in a collaborative way, and go from development to production in a short amount of time. Lastly, the brand needs improved model transparency and interpretation to ensure compliance with legal, data, IT, and marketing teams.
Solution:
Epsilon worked with Dataiku to build a real-time decision engine leveraging Dataiku for model development, workflow, and execution, automation node for job scheduling and monitoring, and API node for integrated services to the various applications for batch and real-time processing.
We integrated this solution into the brand's enterprise CRM and email marketing applications to deliver hyper-personalized email experiences. As each email campaign is generated from the marketing teams, the system calls our environment to return the next best actions for things like product recommendations, offers and promotions, best content, and best subject lines.
This sense and response environment ensures that no two emails are ever the same, and each one is uniquely personalized for every customer profile.
Impact:
- The machine learning test groups have outperformed the control groups by 47% for revenue per email open.
- The machine learning test groups have driven a 20% lift in conversion rates over the control groups for all emails that were opened.
- To date, the program has generated ~31k in net new revenue week over week.
- Created a first of its kind, content optimization delivery system using deep learning and computer vision models in Dataiku.
